Dissertation Channel Coding for Network

CHANNEL CODING FOR NETWORK COMMUNICATION: AN INFORMATION THEORETIC PERSPECTIVE Channel coding helps a communication system to combat noise and interference by adding “redundancy” to the source message. Theoretical fundamentals of channel coding in point-topoint systems have been intensively studied in the research area of information theory, which was proposed by Claude Shannon in his celebrated work in 1948. A set of landmark results have been developed to characterize the performance limitations in terms of the rate and the reliability tradeoff bounds. However, unlike its success in point-to-point systems, information theory has not yielded as rich results in network communication, which has been a key research focus over the past two decades. Due to the limitations posed by some of the key assumptions in classical information theory, network information theory is far from being mature and complete. For example, the classical information theoretic model assumes that communication parameters such as the information rate should be jointly determined by all transmitters and receivers. Communication should be carried out continuously over a long time such that the overhead of communication coordination becomes negligible. The communication channel should be stationary in order for the coding scheme to transform the channel noise randomness into deterministic statistics. These assumptions are valid in a point-to-point system, but they do not permit an extensive application of channel coding in network systems because they have essentially ignored the dynamic nature of network communication. Network systems deal with bursty message transmissions between highly dynamic users. For various reasons, joint determination of key communication parameters before message ii transmission is often infeasible or expensive. Communication channels can often be non-stationary due to the dynamic communication interference generated by the network users. The objective of this work is to extend information theory toward network communication scenarios. We develop new channel coding results, in terms of the communication rate and error performance tradeoff, for several non-classical communication models, in which key assumptions made in classical channel coding are dropped or revised.